Thin film deposition techniques are widely used in the manufacturing of microfeatures to form a coating on a workpiece that closely conforms to the surface topography. The size of the individual components in the devices is constantly decreasing, and the number of layers in the devices is increasing. As a result, the density of components and the aspect ratios of depressions (i.e., the ratio of the depth to the size of the opening) are increasing. The size of workpieces is also increasing to provide more real estate for forming more dies (i.e., chips) on a single workpiece. Many fabricators, for example, are transitioning from 200 mm to 300 mm workpieces, and even larger workpieces will likely be used in the future. Thin film deposition techniques accordingly strive to produce highly uniform conformal layers that cover the sidewalls, bottoms, and corners in deep depressions that have very small openings.
One widely used thin film deposition technique is Chemical Vapor Deposition (CVD). In a CVD system, one or more precursors that are capable of reacting to form a solid thin film are mixed in a gas or vapor state, and then the precursor mixture is presented to the surface of the workpiece. The surface of the workpiece catalyzes the reaction between the precursors to form a solid thin film at the workpiece surface. A common way to catalyze the reaction at the surface of the workpiece is to heat the workpiece to a temperature that causes the reaction.
Although CVD techniques are useful in many applications, they also have several drawbacks. For example, if the precursors are not highly reactive, then a high workpiece temperature is needed to achieve a reasonable deposition rate. Such high temperatures are not typically desirable because heating the workpiece can be detrimental to the structures and other materials already formed on the workpiece. Implanted or doped materials, for example, can migrate within the silicon substrate at higher temperatures. On the other hand, if more reactive precursors are used so that the workpiece temperature can be lower, then reactions may occur prematurely in the gas phase before reaching the substrate. This is undesirable because the film quality and uniformity may suffer, and also because it limits the types of precursors that can be used.
Atomic Layer Deposition (ALD) is another thin film deposition technique. FIGS. 1A and 1B schematically illustrate the basic operation of ALD processes. Referring to FIG. 1A, a layer of gas molecules A, coats the surface of a workpiece W. The layer of Ax molecules is formed by exposing the workpiece W to a precursor gas containing Ax molecules and then purging the chamber with a purge gas to remove excess Ax molecules. This process can form a monolayer of Ax molecules on the surface of the workpiece W because the Ax molecules at the surface are held in place during the purge cycle by physical adsorption forces at moderate temperatures or chemisorption forces at higher temperatures. Referring to FIG. 1B, the layer of Ax molecules is then exposed to another precursor gas containing By molecules. The Ax molecules react with the By molecules to form an extremely thin layer of solid material on the workpiece W. The chamber is then purged again with a purge gas to remove excess By molecules.
FIG. 2 illustrates the stages of one cycle for forming a thin solid layer using ALD techniques. A typical cycle includes (a) exposing the workpiece to the first precursor Ax, (b) purging excess Ax molecules, (c) exposing the workpiece to the second precursor By, and then (d) purging excess By molecules. In actual processing, several cycles are repeated to build a thin film on a workpiece having the desired thickness. For example, each cycle may form a layer having a thickness of approximately 0.5-1.0 Å, and thus it takes approximately 60-120 cycles to form a solid layer having a thickness of approximately 60 Å.
FIG. 3 schematically illustrates a single-wafer ALD reactor 10 having a reaction chamber 20 coupled to a gas supply 30 and a vacuum 40. The reactor 10 also includes a heater 50 that supports the workpiece W and a gas dispenser 60 in the reaction chamber 20. The gas dispenser 60 includes a plenum 62 operably coupled to the gas supply 30 and a distributor plate 70 having a plurality of holes 72. In operation, the heater 50 heats the workpiece W to a desired temperature, and the gas supply 30 selectively injects the first precursor Ax, the purge gas, and the second precursor By as shown above in FIG. 2. The vacuum 40 maintains a negative pressure in the chamber to draw the gases from the gas dispenser 60 across the workpiece W and then through an outlet of the reaction chamber 20.
One drawback of ALD processing is that it has a relatively low throughput compared to CVD techniques. For example, each Ax-purge-By-purge cycle can take several seconds. This results in a total process time of several minutes to form a single thin layer of only 60 Å. In contrast to ALD processing, CVD techniques require only about one minute to form a 60 Å thick layer. The low throughput of existing ALD techniques limits the utility of the technology in its current state because ALD may be a bottleneck in the overall manufacturing process.
Another drawback of ALD and pulsed CVD processing is that it is difficult to control the amount of gas in the chamber at the beginning of a pulse. In ALD and CVD processing, a mass flow controller or a pressure controller typically controls the flow of precursors into the reactor. For example, when a mass flow controller controls the gas flow, the beginning of the pulse can have an undesirably high flow rate. Mass flow controllers are susceptible to producing high initial flow rates because they continue to cause the gas to flow through the gas line even when a final valve in the supply line is closed to prevent the precursor from flowing into the reactor. This causes the pressure in the supply line to continually increase. Moreover, as the flow rate drops when the final valve is closed, the mass flow controller causes an upstream valve to open more. As such, when the final valve opens, the increased pressure in the supply line and the open upstream valve creates a large inrush of precursors into the reaction chamber. This can cause premature reactions between the precursors or remove molecules from the surface of the workpiece W, which can result in a nonuniform surface on the workpiece W. Furthermore, the increased pressure in the supply line can cause the final valve to dispense too much precursor into the reaction chamber, which can make it difficult to purge or otherwise remove the excess precursor from the reaction chamber.
Systems that use pressure controllers to control the gas flow also have drawbacks. For example, the pressure controller typically has a fixed or adjustable orifice in the gas supply line to limit the precursor flow when the pressure drops downstream. The pressure controller, however, cannot control the mass flow rate, and consequently, differences in the downstream pressure can cause different mass flow rates in the gas supply line. Accordingly, too much or too little precursor may be dispensed into the reaction chamber in systems that rely on pressure controllers to modulate the gas flow into the reactor. Thus, there is a need to improve the process of providing precursors to reactors.